Ischemia detection using Isoelectric Energy Function
Introduction
In recent years, epidemiological transition due to obesity, smoking, tobacco use, physical inactivity, an unhealthy diet and harmful use of alcohol has increased the death rate. An estimated 17.5 million people died from cardiovascular diseases (CVDs) in 2012, representing 31% of all global deaths. Of these deaths, an estimated 7.4 million were due to coronary heart disease and 6.7 million were due to stroke. Out of the 16 million deaths under the age of 70 due to non-communicable diseases, 82% for low and middle-income countries and 37% are caused by CVDs. As per prediction of the world health organization (WHO), 75% of world deaths will be due to non-communicable and coronary heart diseases (CHD) by 2030 [1]. These diseases mainly result from atherosclerosis and thrombosis, which can manifest as a coronary ischemic syndrome [2]. CVDs can be prevented by addressing behavioral risk and through management using medicine and early detection. So, currently, the attention of researchers is on establishing the timely and reliable detection of ischemia. An electrocardiogram (ECG) is one of the non-invasive techniques to diagnose the ischemia. An ECG is representative of the P wave (depolarization of the atria), QRS complex (depolarization of the ventricles) and T wave (ventricular repolarisation). Under normal conditions, an ECG carries a predictable duration, direction and amplitude of characteristic points. However, an ischemic ECG has a peculiar appearance, which is indicative of a decrease in terms of the availability of oxygen for cardiac tissues. ST-segment changes are produced by the flow of injury currents that are generated by the voltage gradients between the ischemic and non-ischemic myocardium during the plateau and resting phases of the ventricular action potential. This is manifested as an elevated or depressed ST segment in an ECG [3]. ST elevation usually appears in patients with transmural ischemia or prinzmetal angina while ST depression appears in subendocardal ischemia or stable or unstable angina [4]. A number of methods have been proposed in the last 10–15 years for the detection of ischemic beats and episodes based on digital signal analysis, rule based wavelet transforms and soft computing based algorithms. These include wavelet based entropy [5], [6], a network self-organizing map (sNet-SOM) model [7], neural network (NN) and nonlinear principal component analysis (NLPCA) [8], a back propagation algorithm [9], Karhunen–Loeve transform (KLT) [10], a genetic algorithm [11], hidden Markov models (HMM) [12], machine learning techniques [13], decision tree rules using fuzzy models [14], vectorcardiographic ST-T [15] methods etc. These methods have the potential of a decision support system that can provide good advice for diagnosis. Other methods based on wavelets [16], an ant-miner algorithm [17], kernel density estimation (KDE) and support vector machine (SVM) [18], statistical features [19] and a fuzzy expert system through stochastic global optimization [20] are reported. These methods have a major advantage of interpretation of decisions as compared to black box approaches like neural networks. A Real-time system for the detection of myocardial infarction [21] is implemented. Similarly, one study identifies the ECG morphologies for normal and abnormal beats based on wavelet power spectra using statistical significance [22]. Likewise, a survey of ischemia detection techniques [23] has been performed. During recording of an ECG, the artifacts, i.e. anything other than the muscular activity of the heart, are contaminated on ECG. Critical problems contributing to poor detection and classification of the ST segment in an ECG include baseline wanders, varying ST-T patterns, muscle tremors, high frequency noises, power line interference, etc. The reasons may be the interference of alternating current, a loose electrode connection, patient movements, malfunctioning of the machine etc. [24]. These artifacts have an adverse effect on an ECG and make the automatic delineation of characteristic points more difficult. Besides these problems, it is necessary to detect ischemic episodes when a patient is in a critical care unit (CCU). An experienced cardiologist could easily diagnose the ischemia by just looking at the ECG, even with the presence of artifacts. However, the elimination of artifacts is primarily required to facilitate easy, accurate and automatic detection of the ST segment and classification of ischemic episodes in pathological cases. The recommendation of the American Heart Association (AHA) is to preserve the linear phase, and high pass filters can be designed with a cut-off frequency equal to the fundamental frequency of a heart rate of or lower than a certain threshold, i.e. 0.8 Hz. A number of filters have been designed for the correction of baseline wandering, including FIR and IIR filters. But these filters have the disadvantage that the ECG signal is deformed as the cut-off frequency increases. Cubic spline filters overcome this without the effect of deformation, but these filters make several errors when the sampling rate is low or when the baseline changes suddenly. Adaptive filters determine the signal and adaptively remove the noise uncorrelated with the deterministic signal. The disadvantage of these filters is the distortion of the ST segment [25]. Similarly, for ECG enhancement, the most widely used algorithm is the least mean square adaptive algorithm (LMS). But this algorithm is not able to track the rapidly varying non-stationary ECG signal within each heart beat. Using the method presented in this paper, all the listed problems can be avoided because a wavelet transform is found to be more suitable for a non-stationary ECG signal [26]. Due to the time–frequency resolution capability of wavelet transforms (WT), they were found to be more suitable for removal of noises and artifacts and for the delineation process in ECG records as compared to Fourier transform (FT) and short time Fourier transform methods (STFT) [27]. The general objective of this paper is to propose and validate a simple method for the detection of ischemia based on an isoelectric energy function through accurate pre-processing and detection of basic ECG characteristic points. Our contribution towards the diagnosis of ischemia is twofold. First, the proposed method could provide an interpretation of the results. This is of great importance while designing a device for medical decision support for patients in a critical care unit (CCU) without knowing past references. Second, it involves a direct analysis based on isoelectric energy without the involvement of any complicated calculations. The paper is organized in six sections: Section 1 introduces the ECG, ischemia and related work, Section 2 deals with resources and methods, Section 3 presents the methodology, Section 4 discusses the results, Section 5 makes a comparison with existing methods and Section 6 covers conclusions and the future scope.
Section snippets
European ST-T database
Normal and ischemic (elevated and depressed) ECG records have been taken from the European ST-T database (EDB). EDB records are well characterized digital recordings of ECGs, which are used by most biomedical researchers for validation of their algorithms. The EDB contains 90 ECG signals for two hour recordings with a sampling frequency of 250 Hz per channel over a 10 mV range with 11-bit resolution [28]. The database contains records of two leads acquired from V1, V2, V3, V4 and V5 and MLI and
Methodology
A general schematic diagram is shown in Fig. 1, also representing the methodology involved in the detection of ischemia for proposed method.
Results and discussion
The pre-processing and delineation algorithms are implemented using the wavelet transform toolbox in MATLAB2012a. We have validated the proposed algorithm for 10 representative records of the annotated ESC ST-T database, namely the whole e0104 recording and the first hour of the e0103, e0105, e0108, e0113, e0114, e0147, e0159, e0162 and e0206 recordings. The selected ECG signals contain 20 ischemic ST segment episodes. These records covered both elevated and depressed ST segment cases. These
Comparison with existing methods
This paper proposes a new function, named the isoelectric energy function for detection of ischemia, which does not involve any complex calculations. The performance of the algorithm has been validated on the above mentioned 10 records of the European ST-T database for comparison with existing methods. This database is used as a standard reference for classification of ischemic beats. Specifically, it was considered that every annotated episode in the database contained only ischemic beats. Our
Conclusion and future scope
We have successfully proposed and validated a simple method for the diagnosis of ischemia based on a threshold of isoelectric energy of ST segments in ECG signals. The multi-resolution features of a wavelet transform have been employed for the pre-processing and delineation of ECG characteristic points. The proposed method achieves significantly better results than do the existing methods in the literature. The method does not involve any complex calculations and could also employ additional
Conflict of interest statement
The authors have nothing to declare.
References (49)
- et al.
A methodology for the automated creation of fuzzy expert systems for ischemic and arrhythmic beat classification based on a set of rules obtained by a decision tree
Artif. Intell. Med.
(2007) - et al.
Novel technique for ST-T interval characterization in patients with acute myocardial ischemia
Comput. Biol. Med.
(2014) - et al.
ECG analysis using wavelet transform: application to myocardial ischemia detection
ITBM-RBM
(2003) - et al.(2014)
- et al.
Current methods in electrocardiogram characterization
Comp. Biol. Med.
(2014) - et al.(2012)
- et al.
An ischemia detection method based on artificial neural networks
Artif. Intell. Med.
(2002) - WHO Global Status Report on Noncommunicable Diseases 2014, isbn: 978 92 4 156485 4, available at...
Introduction to Biomedical Instrumentation
(2010)- et al.
AHA/ACCF/HRS recommendations for the standardization and interpretation of the electro-cardiology rhythm society: endorsed by the international society for computerized on clinical cardiology; the American College of Cardiology Foundation; and the Heart American Heart Association Electrocardiography and Arrhythmias Committee, Council
Circulation
(2009)
Electrocardiogram monitoring
Indian J. Anaesth.
Ischemia detection by electrocardiogram in wavelet domain using entropy measure
J. Res. Med. Sci.
Wavelet time entropy, T wave morphology and myocardial ischemia
IEEE Trans. Biomed. Eng.
Ischemia detection with a self organizing map supplemented by supervised learning
IEEE Trans. Neural Netw.
ECG analysis using nonlinear PCA neural networks for ischemia detection
IEEE Trans. Signal Process.
An adaptive back propagation neural network for real-time ischemia episodes detection: development and performance analysis using the European ST-T database
IEEE Trans. Biomed. Eng.
Detection of ST segment deviation episodes in ECG using KLT with an ensemble neural classifier
Physiol. Meas.
Automated ischemic beat classification using genetic algorithm and multi-criteria decision analysis
IEEE Trans. Biomed. Eng.
ST segment analysis using hidden markov model beat segmentation: application to ischemia detection
Comput. Cardiol.
Real-time detection of transient cardiac ischemic episodes from ECG signals
Physiol. Meas.
Rule based classification of ischemic ECG beats using ant-miner
Int. J. Eng. Sci. Technol.
Ischemia episode detection in ECG using kernel density estimation, support vector machine and feature selection
Biomed. Eng. Online
Analysis of electrocardiogram signals of arrhythmia and ischemia using fractal and statistical features
J. Mech. Med. Biol.
A framework for fuzzy expert system creation – application to cardiovascular disease
IEEE Trans. Biomed. Eng.
Cited by (29)
An automated patient-specific ECG beat classification using LSTM-based recurrent neural networks
2023, Biomedical Signal Processing and ControlAn overview on state-of-the-art electrocardiogram signal processing methods: Traditional to AI-based approaches
2023, Expert Systems with ApplicationsThe role of optimal and modified lead systems in electrocardiogram
2022, Advanced Methods in Biomedical Signal Processing and AnalysisTime-frequency approach to ECG classification of myocardial infarction
2020, Computers and Electrical EngineeringCitation Excerpt :It achieved 96.20%, 94.60%, 91.50% and 94.64% test accuracies for classes ST Depression ST Elevation, Healthy and Arrhythmia, respectively. As revealed in Table 8, the accuracy of the studies [5–9,13,14,16,17,19], the sensitivity values of the studies [6–8,14,20], and specificity values of the studies [7,8,20] are slightly higher than that of the proposed method. It is worth to mention that, the proposed method is a four-class study, whereas other studies are two-class or three-class studies.
Ballistocardiogram heart rate detection: Improved methodology based on a three-layer filter
2020, Measurement: Journal of the International Measurement ConfederationCitation Excerpt :Heart rate, an important indicator of cardiovascular activity, refers to the number of heart beats per minute. The resting heart rate of a normal adult varies between 60 and 100 beats/min [1], with an average of 75 beats/min. Clinical studies have shown that heart rate has a significant relationship with cardiovascular disease, as a high resting heart rate suggests an increased the risk of cardiovascular disease [2].
ECG fiducial point extraction using switching Kalman filter
2018, Computer Methods and Programs in BiomedicineCitation Excerpt :Low pass differentiation (LPD) [9], hidden Markov models (HMM) [10–14], partially collapsed Gibbs sampler (PCGS) [15], wavelet transform [16–18], correlation analysis [19,20] and extended Kalman filter (EKF) [21–24] are also used for ECG FP extraction. FP extraction has been used as a preprocessing step in several applications such as detection of fragmented QRS complex [25], mobile health care applications [26], “Selvester QRS scoring” system [27], ischemia detection [28], ECG-based subject identification system [29] and biometric recognition based on fusion of ECG and EEG signals [30,31]. Switching state space models are defined as the combination of HMMs and state space models [32].